Error of MaskDINODecoder #127
Description
src_flatten = []
mask_flatten = []
spatial_shapes = []
for i in range(self.num_feature_levels):
idx=self.num_feature_levels-1-i
bs, c , h, w=x[idx].shape
size_list.append(x[i].shape[-2:])
spatial_shapes.append(x[idx].shape[-2:])
src_flatten.append(self.input_proj[idx](x[idx]).flatten(2).transpose(1, 2))
mask_flatten.append(masks[i].flatten(1))
src_flatten = torch.cat(src_flatten, 1) # bs, \sum{hxw}, c
mask_flatten = torch.cat(mask_flatten, 1) # bs, \sum{hxw}mask_flatten.append(masks[i].flatten(1)) # bug: error, maybe is masks[idx]???????
If my understanding is correct, there appears to be a logical error in the mask processing within the decoder implementation. While this doesn't affect the final results, the implementation logic seems incorrect.
Please correct me if I've misinterpreted the code.